Abstract: A polarity classification approach of web review based on dependency analysis is proposed.Dependencies and part-of-speeches are extracted from web review to construct dependency and part-of-speech pairs，and then classification by polarity dictionary is accomplished.The approach can effectively reduce the complexity of calculation and improve the accuracy of classification，by using dependency and part-of-speech pair and custom polar dictionary.This experiment results show that the approach will obtain higher performance compared with other classification approaches.It is a feasible and valid polarity classification approach of web review.

Key words: dependency analysis, polarity classification, dependency and part-of-speech pair, Web review

摘要： 介绍一种利用句法依存关系对网络评论的极性进行自动分类的方法。通过从评论中提取出依存关系和词性，构成依存关系词性对，并利用自定义的极性词典进行分类，有效地减少计算的复杂度和提高分类的精度。实验表明，该方法相比其他方法取得了较好的分类效果，是一种可行且有效的对评论极性分类的方法。

关键词: 依存关系分析, 极性分类, 依存关系词性对, 网络评论